Kernel 2.6.7 speedstep-piix4 on vaio

Kernel 2.6.7 speedstep-piix4 on vaio

[Ivor Hewitt]

Hi,

Firstly should the speedstep-smi driver be able to work on a Vaio Z600TEK
(Coppermine P3, 440BX chipset)? I can only get speedstep to work on 2.6 by
using the old piix4 code. Has anyone else got speedstep going on one of these?

For me the speedstep-smi driver fails on this machine with the error:
"could not detect low and high frequencies by SMI call"
"could not detect two different speeds -- aborting"

However, I used to have the piix4 driver working on this machine, but this is
not in the 2.6 kernel tree.

So... I updated the old copy of the piix4 driver which used the old gpo_hilo
code to work on kernel 2.6.7. and it now works just fine for me.

The text on the www.poupinou.org site implied that the piix driver was now
obsolete due to the smi driver.

Given that I now have the piix driver working, should it be possible to make
the smi driver work? i.e. is the smi just a better way of doing the same thing
and anything that the old driver did should be possible in the new one?

I'll post my patch to get piix working against 2.6.7, when I get back to a
network connection.

Regards,
Ivor.

PS Is there a searchable archive of this list?

-- 
http://ivor.org - leylandii hedge -- http://ivor.it - technology

SquirrelMail - webmail for nuts.

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Dominik Brodowski]

Hi,

On Thu, Jul 29, 2004 at 03:20:00PM +0100, Ivor Hewitt wrote:
> Hi,
> 
> Firstly should the speedstep-smi driver be able to work on a Vaio Z600TEK
> (Coppermine P3, 440BX chipset)?
Possibly.

> Given that I now have the piix driver working, should it be possible to make
> the smi driver work? i.e. is the smi just a better way of doing the same thing
> and anything that the old driver did should be possible in the new one?

No, the SMI driver does things differently. the -piix4 driver changes an
output signal of the southbridge which might or might not be connected to
the appropriate "intel speedstep logic" to switch speeds. the -smi driver
tries to talk to the BIOS which knows how to switch speeds, but sometimes
the BIOS doesn't understand us, as it only understands an undocumented
language [== undocumented values]...

Therefore, the speedstep-smi driver contains a few tweaks which modify the
"language" a bit: have you tried passing the module options
"smi_port=0xb2 smi_cmd=0x82" yet? On several systems, the smi_cmd is set
wrongly by default.

> PS Is there a searchable archive of this list?
Possibly @ gmane, but I don't know how far back that archive goes...

	Dominik

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Ivor Hewitt]

On Thursday 29 July 2004 21:04, Dominik Brodowski wrote:
> On Thu, Jul 29, 2004 at 03:20:00PM +0100, Ivor Hewitt wrote:
>
> Therefore, the speedstep-smi driver contains a few tweaks which modify the
> "language" a bit: have you tried passing the module options
> "smi_port=0xb2 smi_cmd=0x82" yet? On several systems, the smi_cmd is set
> wrongly by default.
>
Aha, with the right combination of BIOS settings and the smi_cmd=0x82 setting 
the smi driver works on the Vaio Z600.

Had to set BIOS settings as:-
PnP OS: No
CPU OS Control: Auto
CPU Speed: Auto

Many thanks,

-- 
Ivor Hewitt.
http://www.ivor.it - tech | http://www.ivor.org - hedge

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Dave Jones]

On Fri, Jul 30, 2004 at 08:20:49PM +0000, Ivor Hewitt wrote:
 > On Thursday 29 July 2004 21:04, Dominik Brodowski wrote:
 > > On Thu, Jul 29, 2004 at 03:20:00PM +0100, Ivor Hewitt wrote:
 > >
 > > Therefore, the speedstep-smi driver contains a few tweaks which modify the
 > > "language" a bit: have you tried passing the module options
 > > "smi_port=0xb2 smi_cmd=0x82" yet? On several systems, the smi_cmd is set
 > > wrongly by default.
 > >
 > Aha, with the right combination of BIOS settings and the smi_cmd=0x82 setting 
 > the smi driver works on the Vaio Z600.

hmm, I wonder if its worth adding DMI matches for certain laptops
that need different arguments.

 > Had to set BIOS settings as:-
 > PnP OS: No
 > CPU OS Control: Auto
 > CPU Speed: Auto

Though it wouldn't help with this, we'd just have to hope for the best.

		Dave

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Dominik Brodowski]

On Mon, Aug 02, 2004 at 08:31:41PM +0100, Dave Jones wrote:
> On Fri, Jul 30, 2004 at 08:20:49PM +0000, Ivor Hewitt wrote:
>  > On Thursday 29 July 2004 21:04, Dominik Brodowski wrote:
>  > > On Thu, Jul 29, 2004 at 03:20:00PM +0100, Ivor Hewitt wrote:
>  > >
>  > > Therefore, the speedstep-smi driver contains a few tweaks which modify the
>  > > "language" a bit: have you tried passing the module options
>  > > "smi_port=0xb2 smi_cmd=0x82" yet? On several systems, the smi_cmd is set
>  > > wrongly by default.
>  > >
>  > Aha, with the right combination of BIOS settings and the smi_cmd=0x82 setting 
>  > the smi driver works on the Vaio Z600.
> 
> hmm, I wonder if its worth adding DMI matches for certain laptops
> that need different arguments.

I think the DMI table would get big over time... and considering Linus'
disliking of the ~10 entries in asus_hides_smbus PCI quirk, I don't know
what the general consensus on a big (~50? 100?) table is.

	Dominik

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Ivor Hewitt]

On Monday 02 August 2004 19:31, Dave Jones wrote:

> hmm, I wonder if its worth adding DMI matches for certain laptops
> that need different arguments.
>
>  > Had to set BIOS settings as:-
>  > PnP OS: No
>  > CPU OS Control: Auto
>  > CPU Speed: Auto
>
> Though it wouldn't help with this, we'd just have to hope for the best.
>

I think that would be a good thing to do since it is such a common fix 
(changing the smi_cmd).... although I had tried the smi setting before 
juggling the bios settings which obviously would stop the fix working.... 
Although these were the default settings so it would work out of the box 
normally. I'd been tweaking them before to try and fix an annoying sound card 
problem! ho hum. Push one bubble down in the wallpaper and another one pops 
up. :-)

Regards,

-- 
Ivor Hewitt.
http://ivor.org - leylandii hedge -- http://ivor.it - technology

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Dave Jones]

On Mon, Aug 02, 2004 at 10:27:29PM +0200, Dominik Brodowski wrote:
 > On Mon, Aug 02, 2004 at 08:31:41PM +0100, Dave Jones wrote:
 > > On Fri, Jul 30, 2004 at 08:20:49PM +0000, Ivor Hewitt wrote:
 > >  > On Thursday 29 July 2004 21:04, Dominik Brodowski wrote:
 > >  > > On Thu, Jul 29, 2004 at 03:20:00PM +0100, Ivor Hewitt wrote:
 > >  > >
 > >  > > Therefore, the speedstep-smi driver contains a few tweaks which modify the
 > >  > > "language" a bit: have you tried passing the module options
 > >  > > "smi_port=0xb2 smi_cmd=0x82" yet? On several systems, the smi_cmd is set
 > >  > > wrongly by default.
 > >  > >
 > >  > Aha, with the right combination of BIOS settings and the smi_cmd=0x82 setting 
 > >  > the smi driver works on the Vaio Z600.
 > > 
 > > hmm, I wonder if its worth adding DMI matches for certain laptops
 > > that need different arguments.
 > 
 > I think the DMI table would get big over time... and considering Linus'
 > disliking of the ~10 entries in asus_hides_smbus PCI quirk

I'd like to live in a utopia where broken hardware doesn't exist too, but
last time I checked, neither I, nor Linus does 8-)
I've not had any problems merging such entries with him in the past that
I can recall.

 > I don't know what the general consensus on a big (~50? 100?) table is.

I'll bet you can get the more common models in much less entries than that.
Given its now an end of life'd chipset, its not like we'll ever see
new entries once we have the more common ones too.  Plus we'd only
need entries from models that deviate from the defaults, and finally
we'd need input from users, which at times, can be hard to gather anyway.
So I'm sceptical that we'll see 50-100 entries.

		Dave

Re: Kernel 2.6.7 speedstep-piix4 on vaio

[Dominik Brodowski]

On Mon, Aug 02, 2004 at 09:49:21PM +0100, Dave Jones wrote:
> I'd like to live in a utopia where broken hardware doesn't exist too, but
> last time I checked, neither I, nor Linus does 8-)
> I've not had any problems merging such entries with him in the past that
> I can recall.

Excellent. Andrew just pushed the "difficult" patch to Linus again, so I'm
still hopeful :)

>  > I don't know what the general consensus on a big (~50? 100?) table is.
> 
> I'll bet you can get the more common models in much less entries than that.
> Given its now an end of life'd chipset, its not like we'll ever see
> new entries once we have the more common ones too.  Plus we'd only
> need entries from models that deviate from the defaults, and finally
> we'd need input from users, which at times, can be hard to gather anyway.
> So I'm sceptical that we'll see 50-100 entries.

Let's see. Got one affected system here locally...

	Dominik